Capturing and Holding Attention: The Impact of Emotional Words in Rapid Serial Visual Presentation

Transcription

1 C293 Capturing and Holding Attention 1 Capturing and Holding Attention: The Impact of Emotional Words in Rapid Serial Visual Presentation Karen J. Mathewson, Karen M. Arnell, Craig A. Mansfield Department of Psychology Brock University St. Catharines, Ontario, Canada When two masked, to-be-attended targets are presented within approximately half a second of each other, accurate report of the second target (T2) suffers relative to when targets are presented farther apart in time an attentional blink (AB). In the present study the AB was found to be larger when taboo words were presented as a first target (T1), relative to emotionally neutral, negative, or positive words, suggesting that taboo words received preferential attentional processing. Comparable results were also obtained when taboo words were presented as to-be-ignored distractors in single-target RSVP. Arousal, but not valence, ratings of the emotional words predicted accuracy on subsequent targets in both dual and single task RSVP. Recognition memory for taboo words accounted fully for the negative relationships between arousal ratings and accuracy on subsequent targets, suggesting that arousal-triggered changes in attentional allocation influenced encoding of taboo words at the time they were encountered. In a typical rapid serial visual presentation (RSVP) paradigm, stimuli are presented rapidly (approximately 10/second) one at a time in the same central location. Participants are usually able to detect or identify a specified single target with a high degree of accuracy (Raymond, Shapiro, & Arnell, 1992). However, they have difficulty reporting the second of two targets if the second target (T2) is presented within about 500 ms of the first target (T1) an effect known as the attentional blink (AB; Broadbent & Broadbent, 1987; Raymond et al., 1992). No AB is observed if participants are instructed to ignore T1 and report only T2 (Raymond et al., 1992), or if the targets are presented farther apart in time. According to prominent two-stage bottleneck models of the AB (Chun & Potter, 1995; Jolicoeur, 1999), processing a target to the level of identification requires two discrete steps: 1) processing and representation of stimulus features and 2) sustained attention resulting in consolidation of stimulus identity sufficient for recognition or report. 1 While the first step occurs automatically, the second requires substantial attentional resources and requires more time to complete. The time- and attention-consuming nature of stage two means that if T2 arrives before consolidation of T1 is completed, its own consolidation must wait for presently occupied attentional resources to become available (Chun & Potter, 1995; Jolicoeur, 1999). If processing of T1 outlasts T2 s brief representation before it is overwritten by the subsequent stimulus in the RSVP stream, then encoding of T2 will fail, and T2 report accuracy will be reduced. Thus, such AB theories predict that prolonged processing of T1 at the consolidation stage will result in poor accuracy for T2. Despite short stimulus exposures, there is evidence that word stimuli receive semantic analysis in RSVP during stage one processing. For example, T2s that were blinked and unable to be reported were still able to prime semantically associated words presented after the RSVP stream (Shapiro, Driver, Ward, & Sorensen, 1997) and the semantic relationship between words in RSVP streams has been shown to influence target performance (Maki, Frigen, & Paulson, 1997). Barnard, Scott, Taylor, May and Knightly (2004)

2 showed that a to-be-ignored distractor word captured attention and reduced report accuracy for a subsequent target if the distractor word was semantically similar to the target category. Eventrelated brain potentials (ERPs) have also shown fully intact N400s for blinked T2s, suggesting complete semantic analysis of T2s that could not be reported (Luck, Vogel, & Shapiro, 1996). Evidence for semantic activation of targets and distractors in RSVP streams has led researchers to investigate whether emotionally laden stimuli might receive preferential attentional processing when presented as RSVP targets or distractors. There is reason to suspect that emotionally laden words may receive preferential processing in RSVP, as several paradigms have shown evidence for preferential processing for some emotional materials under some conditions. When using clinical populations, research with paradigms such as Stroop (see Williams, Mathews, & MacLeod, 1996 for a review) and dot probe (e.g., MacLeod, Mathews, & Tata, 1986) has demonstrated that clinical participants exhibit a compelling bias to attend to emotional words, particularly if they are consistent with their psychopathology. However, emotionally charged words show attentional effects less reliably in normal controls. In participant samples with no clinical psychopathology, sometimes emotionally charged words appear to be exempt from the usual attentional processing limitations, as they are able to capture attention in paradigms such as inattentional blindness (e.g., Mack & Rock, 1998), dot probe (Mogg, Bradley, De Bono, & Painter, 1997), Stroop (MacKay, Shafto, Taylor, et al., 2004) and digit parity (Aquino & Arnell, in press). Imaging studies (Compton, Banich, Mohanty, et al., 2003; Whalen, Bush, McNally et al., 1998a) and electrophysiological data (e.g., Stormark, Nordby, & Hugdahl, 1995) have also provided evidence that the processing of emotional words may be amplified. However, normal control participants have sometimes failed to show attentional biases to emotionally laden words in paradigms such as Stroop (Gotlib & McCann, 1984), dot probe (MacLeod et al., 1986), visual search (Harris, Pashler, & Coburn, 2004), and digit parity (Harris & Pashler, 2004). Therefore, there is currently no clear consensus as to the conditions under which emotional stimuli affect C293 Capturing and Holding Attention 2 attentional processing in normal individuals, nor about the nature of that effect. RSVP and emotionally charged words Within the AB paradigm emotionally charged words may be presented as T1, as T2, or as a distractor in the RSVP stream. The latter two options have been studied quite extensively. Anderson and Phelps (2001) first observed that healthy control participants produced a smaller AB for arousing, negative T2 words, whereas patients with left or bilateral anterior temporal lobectomies that included removal of the amygdalae showed no reduction relative to non-emotional T2 words. Keil and Ihssen (2004) also showed a reduced AB for negative (and positive) verbs presented as T2s during the AB interval, but only when they were rated as arousing. Emotion theorists agree on two fundamental dimensions that are reflected to varying degrees in all emotions, namely, valence and emotional arousal (e.g., Lang 1995; Lang, Greenwald, Bradley & Hamm, 1993; Russell, 1980; see also Mehrabian s pleasure/arousal/dominance model, 1991, 1997). However, in studies using emotionally-laden words, it is often unclear which of the two major dimensions is most critically related to the observed results. For example, emotional Stroop and subliminal presentation studies have reported that negative information automatically attracts attentional resources (e.g., Dijksterhuis & Aarts, 2003; Ito, Larsen, Smith & Cacioppo, 1998; Pratto & John, 1991; see also Whalen, Rauch, Etcoff, et al., 1998b). However, in a systematic investigation of emotional words as second targets in dual-task RSVP, Anderson (2005) reported that the AB was attenuated when sexual/taboo words were presented as T2. He also provided evidence that the arousal ratings of the words, rather than their valence ratings, word frequency or distinctiveness, represented the critical factor in the AB attenuation. Anderson (2005) suggested that highly arousing T2s require less attention for successful identification and report, and are therefore less vulnerable to the attentional limitations that underlie the AB. Overall, AB studies have shown convincingly that arousing T2 words can overcome, at least in part, the attentional limitations that underlie the AB. In contrast, highly valenced T2 words that are less arousing appear to have no effect on the AB (Anderson, 2005; Keil & Ihssen, 2004).

3 Emotional stimuli have also been presented as tobe-ignored distractors in the RSVP stream while participants search for a single target (e.g., Arnell, Killman, & Fijavz, in press; Barnard, Ramponi, & Battye, 2005; Most, Chun, & Widders, 2005a). By presenting the emotional stimuli at various temporal positions prior to the single target, one can use single target accuracy to estimate the amount of attention given to the emotional stimulus. The results of these studies indicate that the presence of emotional distractors can capture attention at the expense of accuracy for subsequent emotionally neutral targets in effect setting off an involuntary AB (involuntary because the emotional stimulus was not a specified target). Most and colleagues (2005a) have demonstrated that a disturbing photo (e.g., a gory picture of a disgusting scene) presented in the RSVP stream as a to-be-ignored distractor captured participants attention, reducing their accuracy on the subsequent emotionally neutral picture target. Furthermore, this effect was greatest for participants who scored high on a harm avoidance measure. Most, Smith, Cooter, Levy and Zald (in press) showed a similar involuntary AB when erotic pictures were presented as distractors. Barnard and colleagues (2005) also investigated these attentional effects in RSVP, presenting threat words to participants with high or low anxiety levels. Threatening distractors reduced accuracy for subsequent neutral target words, but only in participants characterized by high state and high trait anxiety. Arnell et al. (in press) presented unselected participants with a sad, positive, threatening, taboo, or emotionally neutral word as a to-be-ignored distractor prior to an emotionally neutral target in RSVP. Their results showed that target accuracy was uniformly high when targets were preceded by a neutral, threatening, sad, or positive distractor, but was reduced when a taboo distractor appeared shortly before the target. Furthermore, they found that arousal ratings, but not valence ratings of the emotional words predicted accuracy for subsequent targets. Thus, arousing words presented as T2 attenuate the AB, but arousing words presented as to-beignored distractors appear to initiate an involuntary AB, both suggesting that arousing words receive different attentional processing than less arousing words when presented in RSVP. However, no known studies to date have examined the impact, if C293 Capturing and Holding Attention 3 any, of presenting emotional words as T1 in the AB paradigm. According to two-stage bottleneck models of the AB, T2 cannot be consolidated until T1 has been consolidated and the attentional bottleneck is cleared. If arousing words receive preferential attentional processing, then T1 may receive more attention at the expense of T2, resulting in a larger AB when T1 is emotionally arousing than when it is not. On the other hand, if fewer attentional resources are required to process emotionally arousing words, then T1 may clear the bottleneck more quickly, allowing earlier T2 processing, and resulting in a smaller AB. The Present Study In the present study we examined whether emotionally charged words can influence second target report when the emotional word is voluntarily attended (presented as T1) and when it is not (presented as a to-be-ignored distractor). Including both conditions allowed us to test whether the words that commanded attention in one task were the same words that captured attention in the other task. In the present study we presented emotionally neutral, negative (sadness related), positive, sexual/taboo, and threatening words as T1 and examined the effect on report of a neutral T2 presented at various lags after T1. In a separate condition these same words were presented as tobe-ignored distractors (pseudo-targets) preceding a single to-be-reported target (as in Arnell et al., in press). In addition, to isolate the effects of emotional arousal and valence on second target report, we asked participants to provide ratings of emotional arousal and valence for each of the emotional words. A surprise recognition memory task was also given after the RSVP trials to examine which words had been encoded into memory. The effects of emotional words as T1s and pseudotargets were tested first in Experiment 1 using a within-participants design, followed by a second experiment using a between-subjects design Sexual/taboo words have been shown to attenuate the AB when presented as T2 (Anderson, 2005; Anderson & Phelps, 2001) and to set off an involuntary AB when presented as to-be-ignored distractors in single target RSVP (Arnell et al., in press). If these words do receive preferential attentional processing in RSVP, and T2 must wait until attention is freed from T1 to be consolidated, then we predict that a larger AB will be observed

4 when sexual/taboo words are presented as T1 relative to neutral words. If these words require less attention, and attentional resources are quickly freed for processing of T2, then we expect the AB will be smaller for sexual/taboo T1s. In either case, arousing words should be more likely than other words to be noticed and recognized after a short delay. Indeed, if memory for sexual/taboo words predicts accuracy for targets that follow and recognition performance mediates the relationship between arousal and T2 accuracy, then this will suggest that sexual/taboo words are preferentially encoded at the expense of subsequent neutral targets. Furthermore, we predict that the words that capture attention and set-off an involuntary AB when presented as to-be-ignored distractors in the single-target RSVP paradigm will be the same words that hold attention and increase the AB when presented as T1s in the AB paradigm. Arousal has been intrinsically linked to physiological reactions mediated by the amygdala (Anderson & Phelps, 2001; Canli, Zhao, Brewer, Gabrieli, & Cahill, 2000; Hamann, 2001; Kensinger & Corkin, 2004; Vuilleumier, 2005). When the semantic representations of arousing words are activated, (i.e., at stage 1 processing), the amygdala triggers processes that influence both the perceptual processing of these words and the amount of sustained attention allotted to them (Anderson & Phelps, 2001; Cahill & McGaugh, 1995; Canli et al., 2000; Hamann, 2001; Kensinger & Corkin, 2004; LaBar & Cabeza, 2006; Vuilleumier, 2005), resulting in more complete stage 2 processing. In contrast, words that have high emotional valence but not high arousal show effects mediated by a frontotemporal memory network, not by the amygdala (e.g., Kensinger & Corkin, 2004; LaBar & Cabeza, 2006). As such, we anticipate independent effects for arousal ratings and valence ratings when predicting the magnitude of the AB observed when a given word is presented as T1. Experiment 1 Method Participants Twenty-two Brock University undergraduate students (19 female) participated in this study. They were tested individually in a single session lasting about 1.5 hours, and received partial course credit or a small honorarium for their time. All reported normal or corrected to normal vision C293 Capturing and Holding Attention 4 and English as a first language. The study was reviewed and received approval from the Brock University Research Ethics Board. Design The study was based on two RSVP paradigms, each constructed as a 5 x 6 factorial design. In one paradigm (the AB task), an emotion word (T1), and a target color name (T2), were both embedded in an RSVP stream of emotionally neutral non-color words. There were five kinds of T1 words: neutral, negative, positive, taboo, or one of the distractor words was presented as T1. One of ten target color names appeared in one of six stream positions after the T1 word (1 word later, 2, 3, 4, 5, or 8 words later), corresponding to time lags of 117 ms, 234, 350, 467, 583 and 933 ms, respectively. Whereas all other stimuli were presented in black, T1 was presented in red type. Participants were asked to report whether or not T1 appeared in capital or small letters, and also to identify the T2 color name embedded later in the stream. The levels of each factor varied randomly for each participant, but were constrained so that every possible combination of factors appeared equally often every 60 trials. The AB task consisted of 240 trials. The second paradigm (the capture task) used the same design except that in this task, the emotion word presented as a T1 in the AB task became a tobe-ignored distractor, printed in black uppercase letters like the other stimuli. For each trial, participants were asked only to name the color word embedded in the stream (target), and were not alerted to the presence of the emotion word. The capture task also consisted of 240 total trials. Apparatus and Stimuli All computer stimuli were presented via E-Prime software (Schneider, Eschman, & Zuccolotto, 2002) on a Sony VIAO desktop computer with a 17-inch CRT color monitor. The refresh rate of the monitor was 60 Hz, such that each word required 7 frames for a presentation rate of 117 ms per word. This presentation rate was chosen to produce fairly high levels of T2 performance on both the AB and capture tasks so that any reduction in T2 accuracy could be readily observed. Participants responded by key press, using designated keys on the computer keyboard. For the AB task, words were presented in RSVP format, in which each stimulus appeared in rapid succession in the same place on the screen. There were no blank ISIs between successive items. Each RSVP stream

5 contained 18 items, including a first target (T1; a neutral, negative, positive or taboo word, or one of the words from the distractor set), a target color word (T2), and 16 distractor words. The words subtended about 1.4º of visual angle in height and 3.6º to 7.2º in width at an unfixed binocular viewing distance of approximately 40 cm. All stream items were presented in black, 18-point bold Courier New font on an opaque white screen, with the exception of T1, which was presented in red. On half of the trials in each condition the red T1 word was presented in small letters, and on the other half of the trials in each condition the red T1 word was presented in capital letters. T1 appeared equally often as the 5 th or 8 th item in the stream for each combination of T1 word type and lag. For neutral, negative, positive, and taboo trials, T1 was selected randomly from the appropriate list of 24 words, (see Appendix A) 2 with the constraint that each word was used once every 120 trials. A fifth T1 category consisted of 24 emotionally neutral words randomly drawn from the distractor set (Appendix B) and labeled none. The none category was included to test whether target accuracy would differ for emotionally neutral words that were presented only twice for each task (i.e., words from the neutral category) and for emotionally neutral words that were presented approximately 80 times as distractors (i.e., words from the none category). Each T1 word was presented twice in the AB task. The neutral, negative, positive, and taboo words were adapted from the stimulus set used by Anderson (2005) and were between four and eight letters in length. Across conditions, T1 words were matched for word length (5.13, 5.42, 5.74 and 5.21 letters for neutral, negative, positive, and taboo words respectively) and word frequency 3 (17.83, 23.72, and per million, respectively, Fs < 1; (Kuçera & Francis, 1967). On 20% of the trials the emotion word was absent from the stream, and T1 was a word from the distractor pool that was presented in red. The distractors were 59 neutral valence, low arousal words from four to seven letters in length (Appendix B). For each trial, distractors were chosen randomly from the set without replacement. There were ten possible T2 color words (blue, green, silver, yellow, white, purple, pink, black, brown, and orange) that varied from four to six letters in length. T2 identity was chosen randomly on each trial with the constraint that each color word was used equally often every 20 trials. See Figure 1A for a depiction of the AB task. The stimuli for the capture task were identical to those for the AB task with the exception that the emotion word was now presented as a black distractor (pseudo-target) C293 Capturing and Holding Attention 5 instead of as T1. All stream items were presented in black, uppercase 18-point bold Courier New font on an opaque white screen. See Figure 1B for a depiction of the capture task. Procedure Each participant was asked to perform five different tasks in a single session: 1) a brief questionnaire designed to ascertain the participant s mood in recent weeks, 2) the AB task, 3) the capture task, 4) recognition memory tests for each of these RSVP tasks, where participants identified any T1s they recognized from the AB task and any pseudo-targets they recognized from the capture task, and finally, 5) a rating task in which participants rated T1s/pseudo-targets for their valence and emotional arousal. The order of the capture and AB tasks was counterbalanced across participants. Mood Measure On arrival at the lab, participants were asked to fill out a brief questionnaire (the Hospital Anxiety and Depression Scale; HADS; (Zigmond & Snaith, 1983) that contained seven questions asking about the frequency of depression symptoms and seven questions asking about the frequency of anxiety symptoms. Participants received a score from 0 to 21 for depression symptoms and 0 to 21 for anxiety symptoms. Results from the mood measure are not reported here as they were not reliable across Experiments 1 and 2. AB Task On each trial, participants were instructed to: 1) determine whether the lone red T1 word embedded in the RSVP stream was in capitals or small letters, and 2) identify the color name embedded in the stream following the red word. (The word red was never used as a target in either the AB or capture tasks). Before beginning the experimental trials, participants were shown the 10 color names and informed that the target would always be from this set, and only these responses would be allowed. They were not informed that some of the T1 words would be emotionally laden, beyond a very general warning in the consent form. Before beginning the first task, participants were allowed a few practice trials (average < 10) with the same stimuli as experimental trials. Each trial began with a fixation cross in the center of a blank screen for 500 ms,

6 followed by a 500 ms blank interval before the RSVP stream commenced. Immediately after each stream ended, participants were prompted by the computer to report the whether the red word was presented in capital or small letters. Capital and Small labels were affixed to the keyboard. After making their response, the computer then prompted them to press a key indicating the identity of the target color name. (Color names were attached to 10 keys in the top letter row of the keyboard). Participants were encouraged to guess if they were not certain of the color name, but to try to be accurate. Responses were not speeded. Two seconds after their key press, the fixation cross for the next trial appeared. Capture Task Capture task procedures were the same as those used in the AB task with the exception that on each trial participants were asked to identify the single color target embedded in an RSVP stream, and to ignore all other items in the stream. Participants were not informed that emotion words would be presented on some trials. Recognition Memory Tests After completing both the AB and capture tasks, participants were asked to complete two identical checklists, each containing the 96 words from Appendix A. Participants were told that some of the words had been presented as red targets in the AB task and/or as pseudo-targets in the capture task. They were asked to identify with a check any words they recognized as having seen in the AB task in one list, and in the capture task in another list. In fact, all of the words on the recognition test lists had been presented twice as T1s in the AB task and twice as pseudo-targets in the capture task. To minimize confusion, participants responded with a red pen when identifying T1s they recognized from the AB task (presented in red), and a black pen for identifying black pseudo-targets from the capture task. They were asked to fill out the recognition test list for the most-recently-completed task first. Because the order of task administration was counterbalanced, any difficulty in recalling the most remote (earliest) task would be evenly distributed. Within each list, participants were allowed to go through each list at their own pace, in any order. They could check as many or as few words as they chose. C293 Capturing and Holding Attention 6 Word Ratings Participants were given a short break during instructions for the ratings task. There were 96 trials, in which each of the 96 emotion words from Appendix A (shown as T1s in the AB task and pseudo-targets in the capture task) was presented in random order, balanced across word categories (negative, positive, taboo, neutral). Participants were instructed to rate each word as to its valence (very pleasant to very unpleasant) and arousing quality (very low to very high in emotional arousal) using 7-point Likert scales. With respect to arousal, participants were instructed to consider the magnitude of the reaction they felt when they read a word, regardless of what their reaction was based on. Each word remained on the screen, with a word prompt indicating valence until the participant responded by pressing a number key from 1 to 7 for their valence rating. The same word then remained on the screen but the prompt was changed to arousal. The arousal prompt and the word remained on the screen until the participant pressed a key from 1 to 7 for their arousal rating. Participants were encouraged to consider their subjective reactions to these words carefully and to rate the words using the whole scale. Results Manipulation Check Before beginning the main analyses, we wanted confirmation that our assessment of words as neutral, negative, positive or taboo was also shared by participants. Separate one-way ANOVAs of participants valence and arousal ratings indicated significant effects of word type for each dimension (Valence: F (3, 92) = 84.89, p <.001, η 2 =.74; Arousal: F (3, 92) = 90.53, p <.001, η 2 =.75). Bonferroni-corrected pairwise comparisons in each analysis indicated that words of each word type were rated differently from all other word types for both arousal and valence (all ps <.02). Figure 2 shows a two-factor plot of the mean arousal and valence ratings for each of these words, collapsed across participants. Each word is coded by a symbol representing its type. From this graph it is clear that participants generally agreed with our experimental groupings of words as negative, positive, taboo, and neutral. The none category is not represented in the graph because it consisted of words randomly

7 selected from the 60 neutral distractors, and these words were not rated. AB Task Target Identification Figure 3A shows the mean T2 accuracy (% correct responses) in the AB task for each T1 emotion condition as a function of the lag between T1 and T2. In all experiments, T2 accuracy was calculated for T1-correct trials only, however, the same data patterns were observed for all analyses herein when T2 accuracy was not conditionalized on T1 accuracy. A repeated measures ANOVA was performed on T2 accuracy rates with lag and T1 word type as factors (see Table 1 for all results from this ANOVA). Pairwise comparisons confirmed that T2 accuracy was lower in the T1 taboo condition (M = 65.1% ± 3.1%), relative to each of the other T1 word types (Ms all > 75%), all ps <.001. No other comparisons were significant, all ps >.90. The effect of T1 emotion type differed by lag for the interaction. Pairwise comparisons indicated a significant difference between T2 accuracy on neutral trials and T2 accuracy on taboo trials at lags 3 and 4 (ps <.01), suggesting that the taboo words presented early in the RSVP stream disrupted processing of neutral targets arriving between about ms after the taboo word. However, there were no other significant differences in T2 accuracy for neutral trials versus any other emotion type at any lag (all ps >.05). When task order (AB task performed before or after the capture task) was added as a factor to the above analyses, all original effects remained unchanged, and a significant order by lag interaction was also observed, F (5,100) = 3.06, p <.05, in which the lag effect was larger for participants who performed the AB task second. No interactions including T1 emotion type and order were significant (ps >.34) indicating similar effects of emotion words for both task orders. T1 accuracy was 95.4 % (S.E. =.79) overall, and was unaffected by T1 emotion word-type, lag, or their interaction, all ps >.20. Word Ratings and T2 Identification Relationships among arousal, valence and T2 accuracy in the AB task were examined using the participants arousal and valence ratings for each of the emotion words presented as T1s. For each of the C293 Capturing and Holding Attention 7 96 emotion words, mean valence ratings and mean arousal ratings were calculated by averaging ratings for each word across participants. The average accuracy of T2, collapsed across participants, was also calculated separately for trials using each T1 emotion word. The average arousal and valence ratings for each emotion word were then correlated with mean T2 accuracy on trials where that emotion word was presented as T1. For example, the mean arousal and valence ratings for the word murder were examined with respect to T2 accuracy on trials where murder was presented as T1. Results of these analyses indicated that T1 arousal ratings were associated with reduced T2 accuracy (see Table 2 for all zero-order correlations) even when valence ratings were covaried out (partial r = -.44, t(93) = 4.78, p <.001). Therefore, T1s with higher arousal ratings were associated with lower accuracy for the neutral T2s that followed them. Valence ratings for T1 words did not predict T2 accuracy. However, if the emotional valence of T1 was related to the AB, it may be that valence extremity (very high positive or negative valence compared to neutral valence) would be a better predictor than valence ratings from negative to positive. To examine any potential relationships with valence extremity, the valence rating for each word was subtracted from 4 (the midpoint on the valence scale) and the absolute difference was used as a predictor of T2 performance. Valence extremity did not significantly predict T2 accuracy. However, when arousal ratings and valence extremity values were entered as simultaneous predictors of T2 accuracy, both arousal and valence extremity explained significant unique variability in T2 accuracy (see Table 3). Therefore, once variability due to arousal was accounted for, valence extremity did predict T2 accuracy, but words with more extreme valence were associated with higher, not lower, T2 accuracy. Recognition Hit Rates and T2 Identification T1s with high arousal ratings were associated with poor accuracy for subsequent target information. This may have been because the taboo T1s occupied attentional resources more fully, thereby reducing consolidation of T2. If so, then it is more likely that taboo T1s would be identified in the recognition lists as remembered, relative to other kinds of words. To test this, the recognition

8 hit rate (mean number of checks) for T1 words was compared in each of the conditions of the AB task. The mean number of hits was 3.3 for neutral words, 3.6 for negative words, 4.0 for positive words, and 11.2 for taboo words. A one-way ANOVA confirmed that recognition hits differed significantly across T1 word types, F (3, 92) = 56.02, p <.001. Pairwise comparisons indicated that taboo words were recognized significantly more often than words of any other type, ps <.001, and that the hit rate was the same for all other word types, all ps >.90. The total number of recognition hits each of the 96 T1 words received was summed across participants (maximum = N of 22). The number of hits each word received was then correlated with T2 color accuracy on trials where that word was presented as T1. The correlation revealed a significant negative relationship in which better recognition of a T1 in the AB task was associated with impaired accuracy for the subsequent T2 (see Table 2). There was a positive correlation between the arousal rating of a word and its hit rate, where words rated higher in arousal had more recognition hits. Neither valence ratings nor valence extremity predicted the recognition hits for a word. However, when standardized T1 arousal and valence extremity ratings were entered as simultaneous predictors in a regression on recognition hits for T1 words, both arousal and valence extremity accounted for significant unique variance in recognition, but in opposite directions. T1 words that received higher arousal ratings were better recognized, but T1 words with extreme valence were recognized more poorly (see Table 4). Arousal ratings and recognition hits for T1s were entered together as predictors in a single step in a regression analysis on T2 accuracy. Whereas recognition hits for T1 negatively predicted T2 accuracy even when arousal rating was partialled out (see Table 5), arousal alone did not predict T2 accuracy when variability due to recognition for T1 was removed. This pattern suggests a direct relationship between T1 recognition and T2 accuracy, and an indirect relationship between arousal rating for T1 and T2 accuracy that was mediated by T1 recognition (see Figure 4). Capture Task Target Identification C293 Capturing and Holding Attention 8 Figure 3B illustrates the mean color target accuracy (% correct responses) for each pseudotarget word type (neutral, negative, positive, taboo, and none) as a function of the lag between the pseudo-target and the target (see Table 1 for ANOVA results). Pairwise comparisons showed lower overall accuracy in the taboo condition (M = 81.2%, ± 1.6%) relative to each of the other four pseudo-target conditions (Ms all > 86%), all ps.05, but produced no other significant comparisons. All previous effects were unchanged when task order (the capture task performed before or after the AB task) was added as a factor to the above ANOVA, and there were no main effects or interactions with the task order factor, all ps >.25, indicating similar effects of the emotion words for both task orders. Word Ratings and Target Identification As in the AB task, target accuracy in the capture task was not associated with valence ratings or valence extremity, but higher arousal ratings were associated with poorer identification of subsequent color targets (Table 2). This finding did not change when valence ratings were covaried out (partial r = -.31, t(93) = 3.15, p <.01). As in the AB task, both arousal and valence extremity uniquely predicted target accuracy when examined together in a simultaneous regression. However, words with higher arousal were again associated with lower target accuracy, and those with more extreme valence were associated with higher target accuracy (see Table 3). Recognition Hits and Target Identification A one-way ANOVA on pseudo-target recognition showed that mean recognition hits differed significantly across the pseudo-target types, F (3, 92) = 22.67, p <.001, (M = 3.5 for neutral words, 3.8 for negative words, 4.6 for positive words, and 8.3 for taboo words). Taboo words were recognized significantly more often than words of any other type, all ps <.001, and the number of hits was the same for all other word types, all ps >.50. The total number of recognition hits received by each of the 96 pseudo-targets was summed across participants and correlated with color target accuracy on trials where that word appeared as a pseudo-target. Increased recognition hits for a pseudo-target were associated with significantly impaired accuracy for a subsequent color target

9 (Table 2). Thus, for those pseudo-targets that were later recognized, secondary targets were less likely to be identified during RSVP. This is consistent with the notion that during RSVP attention had been captured by arousing distractors, making it less available for processing or encoding of target information that followed. Again, there was a significant positive relationship between arousal ratings and recognition memory. As in the AB task, there was no relationship between valence or valence extremity and recognition of pseudotargets. However, when variability due to arousal was accounted for, valence extremity also negatively predicted pseudo-target recognition. (see Table 4). Recognition of the pseudo-target was negatively related to target accuracy in simultaneous regression, even when the variability due to arousal rating was removed (see Table 5). However, arousal ratings did not predict target accuracy once recognition memory had been partialled out. Thus, as in the AB task, the analysis is consistent with an entirely mediated model where the relationship between arousal and accuracy for subsequent targets depends on arousing words being processed sufficiently for later recognition (see Figure 4). Cross-task comparisons Two additional results were of interest. A correlational analysis revealed that emotion words that were more likely to be recognized in the AB task were also more likely to be recognized in the capture task, r(94) =.80, p <.001. Also, emotion words that led to T2 report failure in the AB task were roughly the same emotion words that led to target report failure in the capture task, r(94) =.50, p <.001. Overall, the same emotion words were recognized at the expense of subsequent color targets whether these emotion words were presented as targets that required a response in the AB task, or as to-be-ignored distractors in the capture task. Discussion In Experiment 1, we examined the effect of emotional words on the identification of closely following neutral target words in dual-task and single-task RSVP procedures. Whether the emotion word was presented as a first target in the AB task, or as a to-be-ignored distractor in the second task, the same taboo words were found to reduce accuracy for subsequent emotionally neutral targets. Arousal ratings and valence extremity for the C293 Capturing and Holding Attention 9 emotional words predicted accuracy for trailing targets in both RSVP tasks, where poor target accuracy was associated with high arousal and low valence extremity. Taboo words were also recognized more often than other word types in a post-test recognition task. On the basis of analyses showing that recognition memory mediates the relationship between arousal and target accuracy, we suggest that the arousing quality of taboo T1 words and pseudo-targets led to more attention to these words, resulting in poorer identification of second targets. Experiment 2 For both tasks in Experiment 1 the relationship between arousal ratings of the emotion words and accuracy for targets trailing the emotion words was found to be wholly mediated by recognition of the emotion words. Thus, it was prudent to ask in Experiment 2 whether the taboo items identified by participants as remembered were genuinely recognized from the task or whether they had simply been checked because of their titillating nature. In this experiment, the recognition test lists contained foils for each word type as well as the emotion word targets presented during the RSVP task. To rule out stimulus effects, words that were targets for half of the participants were foils for the other half and vice versa. If participants recognized more taboo words from the task they completed, they should identify these words more often than other word types, but show relatively few false alarms to taboo foils presented in the recognition test list. In addition, an effect of word type should be evident in the corrected recognition scores (hits minus false alarms). We also tested the relationship between the corrected recognition scores and subsequent target accuracy. Secondly, the fact that each participant performed both the AB and capture tasks in Experiment 1 may have made it difficult for them to correctly identify the paradigm in which they had seen various words, despite the provision of color-coordinated pens for each recognition test list (red for AB; black for capture). To eliminate any possibility of source confusion in recognition memory between the two paradigms, Experiment 2 presented the AB and capture tasks to two different groups of participants in a between-subjects design. Finally, the none category of T1 words and pseudo-targets was removed given that the neutral

10 and none results did not differ in Experiment 1. It was replaced with a separate category for threat words, as it was unclear whether threat/fear words (e.g., murder) or sexual/taboo words were responsible for the increased AB and capture effects observed for the taboo category in Experiment 1. The small number of threat words (5/24) was removed from the original taboo list and added to a separate threat category. These words were replaced with additional salacious words related to sex or with words that have a taboo quality in the new sexual/taboo condition for Experiment 2. Method Participants Seventy-eight additional undergraduate students (54 female) from Brock University participated in this study. Each was tested individually for over an hour and given credit toward their coursework or a small honorarium. Thirty-eight participants performed the AB task, and forty performed the capture task. Design, Stimuli and Procedure The design, stimuli and procedure of Experiment 2 were identical to those of Experiment 1, with three exceptions. First, participants were assigned in a pseudo-randomized fashion to perform either the AB task or the capture task, according to the order in which they arrived for the study. Second, a single recognition test list was provided to all participants, containing all 240 words used as T1s or pseudotargets in the RSVP tasks. However, in this experiment, half of the participants received 120 of these words (24 from each emotion category) twice in the RSVP task and the remaining 120 words were memory foils that had not been presented in the RSVP task. The other half of the participants for each RSVP task received the opposite set of words as targets and foils (i.e., if murder was presented in the RSVP stream and was therefore a memory target for participant number 1, then it was not presented in the RSVP stream and was therefore a memory foil for participant 2.) Participants provided arousal and valence ratings for all 240 emotion words (targets and foils). Third, the emotion words were now either neutral, negative, positive, physically threatening or sexual/taboo (see Appendix C). C293 Capturing and Holding Attention 10 Results Manipulation Check Figure 5 shows a two-factor plot of the mean arousal and valence ratings for the words from Appendix C. As indicated by the graph and separate one-way ANOVAs, participants valence and arousal ratings confirmed the categorizations used (Valence AB : F (4, 235) = , p <.001, η 2 =.78; Arousal AB : F (4, 235) = , p <.001, η 2 =.71; Valence Cap : F (4, 235) = , p <.001, η 2 =.80; Arousal Cap : F (4, 235) = , p <.001, η 2 =.72). All word types were rated differently from each other, except that negative and threat words did not differ on valence, and positive and negative words did not differ on arousal. AB task Target Identification The mean T2 color target accuracy (% correct responses) for each T1 word type (neutral, negative, positive, sexual/taboo and threatening) is presented in Figure 6A as a function of the lag between T1 and T2 (see Table 1 for ANOVA results). Pairwise comparisons confirmed that overall T2 accuracy was lower in the sexual/taboo condition (M = 64% ±.03 %), as compared to all other conditions, (all Ms > 73%), which did not differ from each other (ps >.90). T1 accuracy was 93 % (S.E. =.01) and was not affected by word type, nor the word type by lag interaction (ps >.30). There was a main effect of lag, F (5, 185) = 4.01, p <.01, η 2 =.10, but no pairwise comparisons were significant. Word Ratings and T2 Identification T1 arousal ratings and valence extremity were again related to T2 accuracy, but valence ratings were not (see Table 2). As in Experiment 1, simultaneous regression revealed that high arousal ratings for T1 words were associated with lower T2 color word accuracy, whereas more extreme valence ratings for T1 words were associated with higher T2 accuracy (Table 3). Recognition Memory and T2 Identification The mean number of recognition hits awarded to words presented in the AB task was 2.7 for neutral words, 3.9 for negative words, 3.8 for positive words, 4.7 for threatening words, and 11.3 for sexual/taboo words. The mean number of hits

11 differed by word type, F (4, 235) = 76.44, p <.001. Sexual/taboo words were more likely to be recognized than neutral or threatening words, (ps <.001) and threat words were more likely to be recognized than neutral words (p <.01). The mean number of foil errors (false alarms) was 0.9 for neutral words, 2.0 for negative words, 2.0 for positive words, 2.0 for threatening words, and 2.4 for sexual/taboo words. The number of false alarms differed by word type, F (4, 235) = 5.80, p <.001, only because neutral words were less likely to be recognized in error than the other word types (ps <.03), which did not differ (all ps >.90). Sexual/taboo foils did not receive more false alarms than other emotion word types. A corrected recognition score for each word was calculated as hits minus false alarms for that word. The mean recognition score for T1 words was 1.77 for neutral words, 1.90 for negative words, 1.75 for positive words, 2.71 for threatening words, and 8.90 for sexual/taboo words, F (4, 235) = 54.96, p <.001, for the ANOVA. Sexual/taboo words were recognized more often than other words (ps <.001) but there were no differences in corrected recognition for the other word types (all ps >.90). As in Experiment 1, higher arousal ratings in the AB task were associated with increased corrected recognition memory scores for T1 words (Table 2), but neither valence nor valence extremity was correlated with corrected recognition scores. However, both T1 arousal and valence extremity ratings accounted for significant unique variance in corrected recognition memory in simultaneous regression analyses, but in opposite directions, consistent with Experiment 1 (see Table 4). Simultaneous regression showed that better corrected recognition for T1 words was associated with poor T2 report accuracy, whereas arousal alone did not predict T2 accuracy once the variability due to recognition for T1 was removed (Table 5). These results follow the pattern found in Experiment 1, i.e., a direct relationship between T1 recognition and T2 accuracy, and an indirect relationship between arousal ratings for T1 and T2 accuracy that was mediated by T1 recognition (see Figure 4). Capture task Target Identification Figure 6B presents the mean color target accuracy (% correct responses) for each pseudo-target type C293 Capturing and Holding Attention 11 (neutral, negative, positive, sexual/taboo and threatening) as a function of the lag between the pseudo-target and target (see Table 1 for ANOVA results). Again sexual/taboo pseudo-targets led to lower overall target accuracy (M = 79% ± 0.2%) relative to all other word-types (Ms > 84%, ps <.04), which did not differ from each other (all ps >.90). Word Ratings and Target Identification Once again higher arousal ratings were correlated with lower color target accuracy. Although valence again did not predict target accuracy, this time valence extremity did (Table 2). As in the capture task from Experiment 1, both arousal and valence extremity accounted for significant amounts of variance in target accuracy in a simultaneous regression analysis, but in opposite directions (see Table 3). Recognition Memory and Target Identification The number of correct recognitions (hits) awarded to pseudo-targets (M = 2.3 for neutral words, 2.0 for negative words, 2.6 for positive words, 2.3 for threatening words and 6.0 for sexual/taboo words) differed across word types, F (4, 235) = 28.75, p <.001. Sexual/taboo words were recognized significantly more often than other word types, (ps <.001), which did not differ from each other, (all ps >.90). The mean number of erroneously identified foils was 1.9 for neutral words, 1.8 for negative words, 2.5 for positive words, 1.8 for threat words, and 2.0 for sexual/taboo words, which did not differ across word types, F (4, 235) = 1.17, p >.50. Corrected recognition scores (hits false alarms) were for 0.4 neutral words, 0.2 for negative words, 0.1 for positive words, 0.5 for threat words, and 4.0 for sexual/taboo words, differing across word type, F (4, 235) = 20.82, p <.001. Taboo/sexual words were better recognized than other word types (ps <.001), which did not differ from each other (all ps >.90). Correlational analyses indicated that corrected recognition scores were positively related to arousal ratings (Table 2). Valence extremity (but not valence) was negatively correlated with corrected recognition. As in Experiment 1, simultaneous regression revealed that higher arousal ratings were associated with better corrected recognition while increased valence extremity was associated with poorer corrected recognition of the pseudo-targets (Table 4).

12 As in Experiment 1, better corrected recognition for pseudo-targets predicted lower target accuracy (Table 2), even when variability due to arousal was partialled out, (partial r = -.21, t(237) = 3.38, p =.001). Unlike Experiment 1, arousal ratings still predicted color target accuracy in the capture task when variance due to corrected recognition for pseudo-targets was removed (Table 5). These results provide support for a partially mediated model where some, but not all, of the relationship between the pseudo-target s arousal rating and accuracy for a subsequent target can be accounted for by increased processing and subsequent recognition of the arousing distractor. Cross-task comparisons Correlational analyses revealed that in Experiment 2, words that were remembered in the AB task were more also likely to be remembered in the capture task, r (238) =.53, p <.001. In addition, T1 words that produced T2 report failures in the AB task were likely to be the same words that impaired target accuracy when presented as distractors in the capture task, r (238) =.38, p <.001. Discussion The results of Experiment 2 were consistent with those of Experiment 1 in that sexual/taboo words presented as T1s or to-be-ignored distractors led to reduced accuracy for subsequent neutral RSVP targets. As in Experiment 1, these sexual/taboo words were also better recognized than other words, and memory for these words mediated the relationship between arousal ratings of the emotional words and report accuracy for subsequent targets. These results suggest that sexual/taboo words are preferentially attended and encoded at the expense of subsequent neutral material in RSVP. Experiment 2 was similar to Experiment 1 with a few exceptions. First, in Experiment 1, it was possible that participants checked taboo words most often simply because they were attention-getting words, and not because they remembered seeing them earlier in the experiment. To remove any potential response bias, the recognition test list was expanded in Experiment 2 to include an equal number of never-seen foils for each word type, and targets and foils were counterbalanced across participants. A less-biased recognition score, (hits minus foil errors), was calculated for each task in C293 Capturing and Holding Attention 12 order to ensure that identification of T1s and distractors represented true recognition for words seen earlier during the task. While participants showed a slight tendency to falsely identify emotional material more often in the AB task, for both tasks the corrected recognition measure replicated the results of Experiment 1, showing that memory for the emotional words was greater for sexual/taboo words. Furthermore, recognition of the emotion words negatively predicted accuracy on targets that followed them. Indeed memory performance was found to fully (AB task) or partially (capture task) mediate the relationship between target accuracy and arousal ratings, just as in Experiment 1. Second, to eliminate any possible confusion or carryover effects, Experiment 2 used a betweensubjects design where participants performed only the AB task or only the capture task. The high degree of similarity in the results from Experiment 1 and Experiment 2 suggest that the results of Experiment 1 were not due to confusion over performing two RSVP tasks and completing two separate recognition test lists. Even though there were two separate groups of participants performing the AB and capture tasks in Experiment 2, we again found that the emotion words that were better recognized and led to reduced accuracy for subsequent targets in the AB task, tended to be the same words that were better recognized and led to reduced accuracy for subsequent targets in the capture task. Finally, because threatening words might be as arousing as taboo words, but due to fear instead of titillation/shock, a new list of physically threatening words was introduced to distinguish fear-inducing, threatening words from sexually-oriented taboo words. For both tasks the results clearly showed that memory was increased for sexual/taboo words, but not for threatening words, and while sexual/taboo words led to reduced accuracy for subsequent RSVP targets, threatening words did not. Thus, although sexual/taboo words appear to receive preferential attentional processing in RSVP, threatening words do not appear to differ from neutral, positive or sadness-related words. General Discussion Overall, there were five goals for the present study. The first was to examine whether the

13 emotional content of the T1 word would modulate the magnitude of the AB a finding that has not yet been reported. In both experiments a larger AB was observed with sexual/taboo T1 words than with neutral, threatening, positive, or negative T1 words which did not differ in the size of the AB. The second goal was to replicate the finding that sexual/taboo words presented as to-be-ignored distractors in RSVP would reduce report accuracy for targets that were presented shortly after an involuntary attentional blink. This finding was replicated in both experiments. Sexual/taboo words set-off an involuntary AB, but other word types did not. The third goal was to examine whether the words that reduced T2 accuracy when presented as T1 in the AB paradigm were the same words that reduced target accuracy when presented as to-beignored distractors in the capture paradigm. A high degree of similarity in the words was found when the same participants performed both RSVP paradigms (Experiment 1), and when different participants performed the two paradigms (Experiment 2), suggesting that the same mechanism may underlie both effects. The fourth goal was to examine whether arousal ratings, valence ratings, or valence extremity could explain why some words were more effective than others at capturing attention at the expense of subsequent targets in RSVP. In both experiments, high arousal ratings were associated with poor accuracy to subsequent targets, but high valence and valence extremity ratings were not indeed, valence extremity was positively associated with target accuracy once arousal was accounted for. The final goal was to ascertain a possible mechanism for the above effects by examining the degree to which the emotional words were recognized in a surprise recognition memory test. In both experiments, sexual/taboo words were recognized more frequently than other words, and better recognition of an emotional word was associated with lower accuracy for subsequent targets. We also observed that memory performance mediated the relationship between arousal ratings and target accuracy, providing a possible mechanism for the present pattern of results. Attentional allocation to emotionally arousing stimuli Sexual/taboo words had a greater detrimental effect on report of subsequent targets than other C293 Capturing and Holding Attention 13 types of emotional words or neutral words, both when presented as T1s in the AB task or as to-beignored distractors in the capture task. These results are consistent with those from other paradigms such as digit parity, inattention blindness, Stroop and dot probe in which emotionally arousing stimuli have been shown to receive preferential attention (e.g., Aquino & Arnell, in press; Mack & Rock, 1998; MacKay et al., 2004; Mogg et al., 1997). As in these other attention paradigms, a benefit of the RSVP task is that any differences in accuracy for subsequent targets across the emotion conditions suggest that emotional arousal exerted its effect on line, during the RSVP task (i.e., at the time of first target appearance), preceding any effects of longer term memorial consolidation of sexual/taboo words (e.g., Sharot & Phelps, 2004). The present findings are also consistent with other studies investigating attention to emotional material in RSVP. Anderson (2005), Keil and Ihssen (2004), and Anderson and Phelps (2001) all observed a reduced AB when arousing words were presented as T2s in RSVP, suggesting that arousing words were able to overcome the attentional deficit that underlies the AB. In addition, Arnell et al. (in press) showed that sexual/taboo words set-off an AB when presented as to-be-ignored distractors, but that threatening, positive, sadness-related, and neutral words did not. Similarly, Most et al. (2005a) and Most et al. (in press) showed that gory disgusting pictures and erotic sexual pictures were both able to create an involuntary AB when presented as to-beignored distractors prior to single targets. Therefore, it appears that emotionally arousing stimuli receive preferential attentional processing in RSVP, whether they are presented as T1, T2 or to-beignored distractors. Anderson (2005) presented two alternative explanations for the reduced AB with sexual/taboo T2s. He reasoned that sexual/taboo T2 words either had a lower threshold for activation that freed them from the attentional limitations that underlie the AB, or they attracted more attentional resources than other, less emotionally arousing words. Anderson showed that the enhanced report of sexual/taboo T2s during the AB interval was independent of T1 accuracy and response latency, suggesting that sexual/taboo T2 words did not attract attention away from T1 processing, but instead required less attention to be reported

14 successfully. This is in contrast to previous studies showing that accuracy for emotionally neutral T2s decreases with increasing T1 latency (see Jolicoeur, 1999). While Anderson (2005) may be correct that processing of sexual/taboo words in RSVP requires less attention, our results suggest that at least in certain circumstances, sexual/taboo RSVP words actually receive more attentional processing, leading to reduced accuracy for trailing targets. One way to account for the apparent divergence between our interpretation of sexual/taboo words receiving greater attentional resources, and Anderson s (2005) interpretation that sexual/taboo words require less attention, is to examine two types of attentional processing (earlier attentional orienting, and later sustained attention) in the AB. Data from multiple studies support the notion that both orienting and sustained attention are involved in the appraisal of novel stimuli (e.g., Most, Scholl, Clifford, & Simons, 2005b; Scherer, 2001; Scherer, Dan, & Flykt, 2006; Schimmack, 2005). By itself, transient orienting to a novel stimulus produces a representation that is too fragmentary to form the basis of a conscious perception. For a visual stimulus to enter conscious awareness, a reiterative process of interpretation and reinterpretation of the stimulus (i.e., sustained attention; Most et al., 2005b) or an initial relevance check followed by multiple appraisals (Scherer, 2001) is necessary. However, the initial orienting is important because the preconscious information gleaned from this stage determines how much sustained attention will be subsequently allocated to the stimulus (e.g., Morris, Öhman & Dolan, 1998; Whalen et al., 1998a). Suppose that sexual/taboo words generate both increased attentional orienting (during stage 1 processing) and increased sustained attention (during stage 2 processing), relative to other words. If so, changing the temporal position of the sexual/taboo word in RSVP could highlight different attentional effects of such words because it would alter the temporal location at which attentional resources are concentrated. When sexual/taboo words are presented as T1s, both stage 1 attentional orienting and stage 2 sustained attention to these words may be enhanced, relative to neutral T1s. While increased stage 1 attentional orienting to a sexual/taboo T1 is unlikely to have C293 Capturing and Holding Attention 14 consequences for an upcoming neutral target, increased stage 2 sustained attention to such a T1 could impair the processing of the second target by appropriating a greater share of available attentional resources, resulting in greater blinking of the emotionally neutral T2. This is consistent with the idea that AB duration reflects the amount of processing allotted to a verbal T1 (Olson, Chun & Anderson, 1991; Hommel & Doeller, 2005), and with resource sharing accounts of the AB (e.g., Shapiro, Schmitz, Martens, Hommel & Schnitzler, 2006). Similarly, stage 1 attentional orienting is likely to be faster and more intense for a sexual/taboo T2 word than a neutral T2. When the strength of its stage 1 attentional orienting is increased, the representation for a sexual/taboo T2 may be able to survive the postponement of its own stage 2 processing until T1 s processing is completed. Therefore, the AB may be larger when emotionally arousing words are presented as distractors or T1s, due to prolonged stage 2 attention allotted to these distractors or T1s, but it can be smaller when the same words are presented as T2s, due to enhanced stage 1 orienting to T2s that allows T2 representations to survive the normal wait while T1 processing is completed. Proposing enhanced stage 1 attentional orienting for sexual/taboo words is consistent with wellknown physiological evidence on the role of the amygdala in mediating arousal (e.g., Kensinger & Corkin, 2004; LeDoux, 2000; LaBar & Phelps, 1998; Cahill & McGaugh, 1998) and with evidence showing that amygdalar coding of emotional stimuli modulates sensory cortex and top-down, frontoparietal attentional circuits, resulting in perceptual enhancement of these stimuli (Vuilleumier, 2005). Thus, because of amygdalar activity and the allocation of additional processing resources, perceptual representations of emotionally arousing or sexual/taboo words may indeed burn brighter and longer (Anderson, 2005) than surrounding neutral words. Kensinger and Corkin (2004) have shown that emotionally arousing words can influence perceptual processing even under divided attention conditions, suggesting that the perceptionand memory-enhancing activity of the amygdala is relatively automatic (see also Whalen et al., 1998b). When sensory processing is enhanced, the stimulus is preferentially selected for early perceptual analysis, which also facilitates its stage 2

15 identification and consolidation in working memory (Keil, Ihssen & Heim, 2006). Subsequent allocation of attention has been shown to improve the discriminability of attended stimuli and accelerate the rate at which they are processed (e.g., Canli et al., 2000; Carrasco & McElree, 2001; Hamann, Ely, Grafton, & Kilts, 1999). Therefore, regardless of the position of the sexual/taboo word in the stream, its perceptual representation and salience should be enhanced automatically. When sexual/taboo words are presented as T1s, this enhancement would be accompanied by an increase in sustained attention that would lead to a greater AB. However, when sexual/taboo words are presented as T2s, this enhancement would allow fragile stage1 representations to outlast T1 processing, resulting in an attenuated AB. Efficacy of sexual/taboo words It has been suggested that even when attentional effects of emotionally arousing words may be found, they are very short-lived. In a digit parity judgment task, Harris and Pashler (2004) showed that the increased RTs observed when digits were paired with threatening incidental emotional distractor words habituated after a single trial. However, in the present study sexual/taboo words were presented 48 times, and each emotional word was presented twice in each paradigm of each experiment. Nonetheless, the detrimental effect of arousing words on subsequent target accuracy was significant in both paradigms in both experiments, and equal in magnitude for the paradigm done first and the paradigm done second in Experiment 1. Thus, the attentional system s preferential response to sexual/taboo words was maintained across a comparatively long period of time. MacKay et al. (2004) have pointed out that repeating a taboo word produces two separate effects: habituation to the word itself, and a more general attenuation in the surprise associated with encountering explicit, taboo words in a laboratory study. Since the unexpectedness of taboo/sexual distractors should wear off with repeated presentations, the longevity of their effects may have less to do with the general surprise of seeing such words in a laboratory setting and more to do with the strongly arousing nature of the words themselves. Indeed, recent results from our laboratory suggest that the sexual/taboo words used C293 Capturing and Holding Attention 15 in the present study are capable of capturing attention when presented as distractors across at least nine presentations of twelve different words, even when participants were shown all of the pseudo-target distractors in advance and instructed to ignore them (Arnell et al., in press). The discrepancy between the apparent ability of sexual/taboo words to attract/maintain attentional resources over several hundred trials (Aquino & Arnell, in press; Arnell et al., in press; MacKay et al., 2004) and the relative inability of other highpriority words to attract/maintain attention (Aquino & Arnell, in press; Arnell et al., in press; Harris et al., 2004) may be partly explained by the fact that sexual/taboo words generate heightened emotional arousal, whereas words that may be salient because they are important or negative (but not particularly arousing) do this to a much lesser degree. Arousal and valence extremity In both experiments we observed that second target accuracy was negatively correlated with arousal ratings, but positively correlated with valence extremity values for emotional words once variance attributable to arousal was accounted for. Thus, arousal ratings and valence extremity with arousal partialled out were related to second target report in opposite directions. Finding that high valence extremity improved second target accuracy was unanticipated. To account for these relationships, we suggest the following: Correlations between arousal and valence extremity were high in all emotion conditions but sexual/taboo. Therefore, when arousal was partialled from valence extremity, much of the variability due to valence extremity was removed from all conditions other than sexual/taboo. In the sexual/taboo condition, many of the sexual/taboo words were high in arousal but low to moderate in valence extremity, including the words found to be most effective at reducing accuracy for subsequent targets (e.g., penis, orgy). Thus, titillating sexual words were highly arousing and more effective than other words at capturing attention, yet they happened to have low valence extremity scores and both of these facts were reflected in the analyses. In contrast to sexual/taboo words, however, highly-valenced words such as positive, negative, and threatening words were poorly recognized on the surprise memory test. Thus, words with higher valence extremity ratings did not show reduced

16 accuracy for subsequent targets generally, and were not preferentially encoded into memory. Only the sexual/taboo words were effective at reducing target accuracy at all. When data from the various emotion conditions were examined separately, the sexual/taboo condition showed significant relationships between second target accuracy and arousal and valence extremity as above, but this was not true for negative, positive, neutral or threat conditions where no relationships were observed. Therefore, outside of the sexual/taboo condition, valence extremity and arousal did not influence target accuracy. This is likely because there was little variability in target accuracy in the other conditions, and as such there was no meaningful variability to predict. It does not appear that the absence of these relationships was due to the high intercorrelations between arousal and valence extremity that existed in all conditions except for sexual/taboo. For the positive, negative, threat and neutral conditions, target accuracy was not predicted by arousal or valence extremity even in the zero-order correlations in which the interrelationship between arousal and valence extremity was irrelevant. Why might emotional arousal show meaningful relationships with accuracy for subsequent targets and recognition memory, where valence and valence extremity do not? First, converging evidence suggests that the two main dimensions of emotional stimuli (arousal and valence) may be supported by different neural networks. Kensinger and Corkin (2004) have shown that distinct neural routes may serve as substrates for the emotionally arousing aspects of words versus their emotional valence. In their study, arousing words elicited greater activation in an amygdalar-hippocampal circuit, whereas negative words elicited greater activation in a prefrontal cortex-hippocampal circuit associated with controlled processes. Similarly, Dillon and colleagues (Dillon, Cooper, Grent- t- Jong, Woldorff & LaBar, 2006) reported a topographical dissociation at the scalp between ERP indices of arousal and semantic cohesion. In support of these views, Maljkovic and Martini (2005) have described independent effects of arousal and valence in terms of the rates at which visual information from photographs accumulates in short term memory. C293 Capturing and Holding Attention 16 Second, sexual/taboo words appear to strongly manipulate arousal, providing perhaps the greatest potential to capture and hold attention. By invoking an initial reaction of shock, embarrassment or heightened interest, these words are highly likely to invoke heightened reiterative perceptual processing and amygdalar-hippocampal binding responses (Anderson, 2005; Anderson & Phelps, 2001; Kensinger & Corkin, 2004; MacKay et al., 2004). The present pattern of results suggests the influence of arousal and activation of the amygdalarhippocampal circuit in response to sexual/taboo words during RSVP target search. A proposed mechanism for poor report of secondary targets Sexual/taboo words were more likely than other words to disrupt T2 report. In two experiments, the negative relationship between arousal ratings and target report accuracy was fully accounted for by recognition performance in both the AB and capture paradigms. We posit that enhanced stage 2 attentional processes supporting consolidation of first targets underlie the improved memory performance. The enhanced attention could be engaged via the modulation of additional brain regions (e.g., parietal) by the amygdalarhippocampal circuit as suggested by Vuilleumier (2005). It appears that the presence of sexual/taboo words did not merely shock the system, but may have triggered the attentional processes conducive to better stage 2 consolidation, that ultimately led to better recognition of these words. This interpretation is consistent with the results and model of MacKay et al. (2004). Using taboo words in Stroop and other tasks, MacKay et al. (2004) demonstrated that the activation of highly arousing meanings facilitated attention to the taboo words and the binding of meaning to other contextual features of the words such as font, color or location. To explain this effect, MacKay et al. (2004), posited that these words produced emotional reactions that engaged the amygdala, which in turn influenced the hippocampus to bind the arousing stimuli to their context. Facilitated binding of taboo words led to superior memory for the words and their attributes (MacKay et al., 2004). Thus the results from the surprise recognition task used here are consistent with these findings. Although the data are correlational, and more than one pattern of paths could explain the

17 interrelationship between target accuracy, arousal ratings, and recognition performance, the present model provides a plausible mechanism to explain the ability of sexual/taboo words to disrupt second target accuracy in RSVP. While arousal ratings did predict memory performance and accuracy for subsequent targets, it is also worth noting that sometimes words with similar arousal ratings had quite different recognition and/or target accuracy scores. Thus, there is likely more to the character of the words that led to particular capture effects than simply arousal. In Experiment 2 we observed that taboo/sexual words produced stronger effects than threat words but why some taboo/sexual words were more effective than others is not presently known. It was not simply the case that the most vulgar of the taboo/sexual words were the most effective. Indeed, some of the sexual/taboo words that were most successful at disrupting T2 report (e.g., orgasm, penis) were not vulgar. Processing of T1 It is clear that the font decision in the AB task could have been completed efficiently based on perceptual information alone, without necessarily reading any of the T1 words. If participants tried to inhibit reading and concentrate on the T1 fonts, then sexual/taboo T1s apparently derailed this strategy, because their meanings were encoded and remembered. If more attention was devoted to sexual/taboo T1s, and this facilitated greater binding of T1 features and superior memory for T1 words, as suggested by MacKay et al. (2004), then one might have expected font accuracy to increase for sexual/taboo T1s relative to other emotion categories. Alternatively, one could conceptualize the T1 font task as an attentional dilemma similar to that of a Stroop task in which participants inhibited focusing on the semantics of the word while attending to whether T1 words were presented in capitals or small letters. If there were greater interference for sexual/taboo words during the T1 task, one might have expected font-decision accuracy to suffer when the meaning of sexual/taboo T1s was attended to. Neither increased T1 accuracy nor decreased T1 accuracy was observed for sexual/taboo T1s, however, as T1 accuracy did not differ across the emotion word categories. Given the high T1 accuracy rate, it is possible that a ceiling effect simply obscured any differences in C293 Capturing and Holding Attention 17 T1 accuracy across the emotion conditions. It is also possible that sexual/taboo T1s did interfere with the T1 font task, but that this interference simply resulted in longer, not less accurate, processing of T1 a finding that could not be observed given the unspeeded nature of the T1 task. Jolicoeur (1999) has shown that when T1 requires an on-line, speeded response, the AB is larger on trials where the T1 response is slow relative to trials where the T1 response is faster, even when T1 accuracy rates are similar. Therefore, it is possible that increased interference was observed between the semantic and font information for sexual/taboo T1s, but that the font decision was performed as accurately just more slowly. In fact, any factor that increased processing time for T1s could lead to delays in their stage 2 attentional consolidation that would in turn impair stage 2 consolidation of T2 (e.g., Chun & Potter, 1995, Jolicoeur, 1999). If Stroop-like interference between the font and meaning of a sexual/taboo T1 were indeed responsible for impaired T2 consolidation, it might be possible to reduce the interference and the increased AB with sexual/taboo T1s by asking participants to report the identity of T1 (a task that would require the use of semantic processing, instead of making the orthogonal font decision). Deciding on the potential for Stroop-like interference in the T1 task is important, as the enhanced AB with sexual/taboo T1s could be due to greater interference during T1 processing rather than greater stage 2 attention as we propose. An interference explanation seems less likely, though, given that similar results were found when sexual words were presented as to-be-ignored distractors, and these distractors required no orthogonal task. Also, memory was better for sexual/taboo words, which one might not expect if their word meanings were inhibited. However, further research will be needed to fully decide between the Stroop explanation and the sustained attention explanation advanced here. Conclusions When arousing sexual/taboo words were presented as T1, a larger AB was observed. These same words were also able to capture attention involuntarily, disrupting accuracy to subsequent targets even when they were presented as to-beignored distractors. The relationship between the arousal value of the emotional word and subsequent

18 target report accuracy was mediated by recognition performance for these first target words, suggesting that emotional words received preferential stage-2 processing at the expense of subsequent targets. Superior memory for sexual/taboo words and reduced accuracy for targets that follow them may depend on the automatic activation of the amygdalar-hippocampal circuit. References Anderson, A. K. (2005). Affective influences on the attentional dynamics supporting awareness. Journal of Experimental Psychology: General, 134(2), Anderson, A. K., & Phelps, E. A. (2001). Lesions of the human amygdala impair enhanced perception of emotionally salient events. Nature, 411(6835), Aquino, J., & Arnell, K. M. (in press). Attention and the processing of emotional words: Dissociating the effects of arousal words. Psychonomic Bulletin & Review. Arnell, K. M., Killman, K., & Fijavz, D. (in press). Blinded by emotion: Target misses follow attention capture by arousing distractors in RSVP. Emotion. Barnard, P. J., Ramponi, C., & Battye, G. (2005). Anxiety and the deployment of visual attention over time. Visual Cognition, 12, Barnard, P. J., Scott, S., Taylor, J., May, J., & Knightley, W. (2004). Paying attention to meaning. Psychological Science, 15, Broadbent, D., & Broadbent, M. (1987). Anxiety and attentional bias: State and trait. Cognition and Emotion, 2, Cahill, L., & McGaugh, J. L. (1995). A novel demonstration of enhanced memory associated with emotional arousal. Consciousness & Cognition, 4, Cahill, L., & McGaugh, J. L. (1998). Mechanisms of emotional arousal and lasting declarative memory. Trends in Neurosciences, 21, Canli, T., Zhao, Z., Brewer, J., Gabrieli, J. D. E., & Cahill, L. (2000). Event-related activation in the human amygdala associates with later memory for individual emotional experience. Journal of Neuroscience, 20, RC Carrasco, M., & McElree, B. (2001). Covert attention accelerates the rate of visual information processing. Proceedings of the National Academy of Sciences, 98(9), C293 Capturing and Holding Attention 18 Chun, M. M., & Potter, M. C. (1995). A two-stage model for multiple target detection in rapid serial visual presentation. Journal of Experimental Psychology: Human Perception and Performance, 21, Compton, R. J., Banich, M. T., Mohanty, A., Milham, M. P., Herrington, J., Miller, G. A., Scalf, P. E., Webb, A., & Heller, W. (2003). Paying attention to emotion: An fmri investigation of cognitive and emotional Stroop tasks. Cognitive, Affective and Behavioral Neuroscience, 3(2), Dillon, D. G., Cooper, J. J., Grent -'t-jong, T., Woldorff, M. G., & LaBar, K. S. (2006). Dissociation of event-related potentials indexing arousal and semantic cohesion during emotional word encoding. Brain and Cognition., 62, Dijksterhuis, A., & Aarts, H. (2003). On wildebeests and humans: The preferential detection of negative stimuli. Psychological Science, 14(1), Gotlib, I. H., & McCann, C. D. (1984). Construct accessibility and depression: An examination of cognitive and affective factors. Journal of Personality and Social Psychology, 47, Hamann, S. B. (2001). Cognitive and neural mechanisms of emotional memory. Trends in Cognitive Sciences, 5, Hamann, S. B., Ely, T. D., Grafton, S. T., & Kilts, C. D. (1999). Amygdala activity related to enhanced memory for pleasant and aversive stimuli. Nature Neuroscience, 2(3), Harris, C. R., & Pashler, H. (2004). Attention and the processing of emotional words and names: Not so special after all. Psychological Science, 15(3), Harris, C. R., Pashler, H., & Coburn, N. (2004). Moray revisited: High-priority affective stimuli and visual search. Quarterly Journal of Experimental Psychology, 57A(1), Hommel, B., & Doeller, C. F. (2005). Selection and consolidation of objects and actions. Psychological Research, 69, Ito, T. A., Larsen, J. T., Smith, N. K., & Cacioppo, J. T. (1998). Negative information weighs more heavily on the brain: The negativity bias in evaluative categorizations. Journal of Personality and Social Psychology, 75,

19 Jolicoeur, P. (1999). Dual-task interference and visual encoding. Journal of Experimental Psychology: Human Perception and Performance, 25, Keil, A., & Ihssen, N. (2004). Identification facilitation for emotionally arousing verbs during the attentional blink. Emotion, 4(1), Keil, A., Ihssen, N., & Heim, S. (2006). Early cortical facilitation for emotionally arousing targets during the attentional blink. BMC Biology, 4(23). Kensinger, E. A., & Corkin, S. (2004). Two routes to emotional memory: Distinct neural processes for valence and arousal. Proceedings of the National Academy of Sciences, 101, Kuçera, H., & Francis, W. (1967). Computational analysis of present-day American English. Providence, RI: Brown. LaBar, K. S., & Cabeza, R. (2006). Cognitive neuroscience of emotional memory. Nature Reviews Neuroscience, 7, LaBar, K. S., & Phelps, E. A. (1998). Arousalmediated memory consolidation: Role of the medial temporal lobe in humans. Psychological Science, 9, Lang, P. J. (1995). The emotion probe: Studies of motivation and attention. American Psychologist, 50(5), Lang, P. J., Greenwald, M. K., Bradley, M. M., & Hamm, A. O. (1993). Looking at pictures: Affective, facial, visceral and behavioral reactions. Psychophysiology, 30, LeDoux, J. E. (2000). Emotion circuits in the brain. Annual Review of Neuroscience, 23, Luck, S. J., Vogel, E. K., & Shapiro, B. (1996). Word meanings can be accessed but not reported during the attentional blink. Nature, 383, Mack, A., & Rock, I. (1998). Inattentional blindness. Cambridge, MA: MIT Press. MacKay, D. G., Shafto, M., Taylor, J. K., Marian, D. E., Abrams, L., & Dyer, J. R. (2004). Relations between emotion, memory and attention: Evidence from taboo Stroop, lexical decision and immediate memory tasks. Memory and Cognition, 32(3), MacLeod, C., Mathews, A., & Tata, P. (1986). Attentional bias in emotional disorders. Journal of Abnormal Psychology, 95, Maki, W. S., Frigen, K., & Paulson, K. (1997). Associative priming by targets and distractors C293 Capturing and Holding Attention 19 during rapid serial visual presentation: Does word meaning survive the attentional blink? Journal of Experimental Psychology: Human Perception and Performance, 23(4), Maljkovic, V., & Martini, P. (2005). Short-term memory for scenes with affective content. Journal of Vision, 5, Mehrabian, A. (1991). Outline of a general emotion-based theory of temperament. In J. Strelau & A. Angleitner (Eds.), Explorations in temperament: International perspectives on theory and measurement. Perspectives on individual differences. New York: Plenum Mehrabian, A. (1997). Comparison of the PAD and PANAS as models for describing emotions and for differentiating anxiety from depression. Journal of Psychopathology and Behavioral Assessment, 19, Mogg, K., Bradley, B. P., De Bono, J., & Painter, M. (1997). Timecourse of attentional bias for threat information in non-clinical anxiety. Behaviour Research and Therapy, 35, Morris, J. S., Öhman, A. & Dolan, R. J. (1998). Conscious and unconscious emotional learning in the human amygdala. Nature, 393, Most, S. B., Chun, M. M., & Widders, D. M. (2005a). Attentional rubber-necking: Cognitive control and personality in emotion-induced blindness. Psychonomic Bulletin and Review, 12, Most, S. B., Scholl, B. J., Clifford, E. R., & Simons, D. J. (2005b). What you see is what you set: Sustained inattentional blindness and the capture of awareness. Psychological Review, 112, Most, S. B., Smith, S. D., Cooter, A. B., Levy, B. N. & Zald, D. H. (in press). The naked truth: Appetitive, arousing distractors impair rapid target perception. Cognition and Emotion. Olson, I. R., Chun, M. M., & Anderson, A. K. (2001). Effects of phonological length on the attentional blink for words. Journal of Experimental Psychology: Human Perception and Performance, 27, Pratto, F. & John, O. P. (1991). Automatic vigilance: The attention-grabbing power of negative social information. Journal of Personality and Social Psychology, 61,

20 Raymond, J. E., Shapiro, K. L., & Arnell, K. M. (1992). Temporary suppression of visual processing in an RSVP task: An attentional blink? Journal of Experimental Psychology: Human Perception and Performance, 18, Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology, 39, Scherer, K. R. (2001). Appraisal considered as a process of multilevel sequential checking. In K. R. Scherer & A. Schorr & T. Johnstone (Eds.), Appraisal processes in emotion (pp ). Oxford, England: Oxford University Press. Scherer, K. R., Dan, E. S., & Flykt, A. (2006). What determines a feeling's position in affective space? A case for appraisal. Cognition and Emotion, 20, Schimmack, U. (2005). Attentional interference effects of emotional pictures: Threat, negativity or arousal? Emotion, 5, Schneider, W., Eschman, A., & Zuccolotto, A. (2002). E-Prime User's Guide. Pittsburgh, PA: Psychology Software Tools, Inc. Shapiro, K. L., Driver, J., Ward, R., & Sorensen, R. E. (1997). Priming of the attentional blink: A failure to extract visual tokens but not visual types. Psychological Science, 8(2), Shapiro, K. L., Schmitz, F., Martens, S., Hommel, B., & Schnitzler, A. (2006). Resource sharing in the attentional blink. NeuroReport, 17, Sharot, T. & Phelps, E. A. (2004). How arousal modulates memory: Disentangling the effects of attention and retention. Cognitive, Affective, and Behavioral Neuroscience, 4, Stormark, K. M., Nordby, H., & Hugdahl, K. (1995). Attentional shifts to emotionally charged cues: Behavioral and ERP data. Cognition and Emotion, 9(5), Vuilleumier, P. (2005). How brains beware: neural mechanisms of emotional attention. Trends in Cognitive Sciences, 9, Whalen, P. J., Bush, G., McNally, R. J., Wilhelm, S., McInerney, S. C., Jenike, M. A., & Rauch, S. L. (1998a). The emotional counting Stroop paradigm: A functional magnetic resonance imaging probe of the anterior cingulate affective division. Biological Psychiatry, 44, Whalen, P. J., Rauch, S. L., Etcoff, N. L., McInerney, S. C., Lee, M. B., & Jenike, M. A. (1998b). Masked presentations of emotional facial C293 Capturing and Holding Attention 20 expressions modulate amygdala activity without explicit knowledge. Journal of Neuroscience, 18, Williams, J. M. G., Mathews, A., & MacLeod, C. (1996). The emotional Stroop task and psychopathology. Psychological Bulletin, 120(1), Zigmond, A. S., & Snaith, R. P. (1983). The hospital anxiety and depression scale. Acta Psychiatrica Scandinavica, 67, Footnotes 1. It should be noted that by stage-2 consolidation of first or second targets in AB tasks we refer to the on-line attentional binding by which stage-1 representations become bound together and encoded into working memory (see Chun & Potter, 1995; Jolicoeur, 1998). This differs from how the term consolidation is often used in the memory literature, where it refers to the slower, involuntary instantiation in long term memory that takes place after encoding is complete. 2. For lack of a better term this category that contains sexual words, curse words, and threat words has been labeled taboo in Experiment 1, and sexual/taboo in Experiment 2. While not all words in this list are taboo, the more obvious label arousing was avoided for this category given that it would be easily confused with the arousal ratings discussed in each of the experiments. 3. It should be noted that valid word frequency estimates for several of the taboo words were impossible to obtain. The frequency estimates used here for the taboo condition are based on those few taboo words that are likely to have valid estimates. However, word frequency is unlikely to account for the enhanced attentional processing that taboo words received. Valid word frequency counts were available for all words in the other three emotion conditions (neutral, positive, and negative) and varied widely within each condition. However, AB size and attentional capture effects did not vary with word frequency. Author Notes This work was made possible through grants to the second author from the Natural Sciences and Engineering Research Council (NSERC), the Canadian Foundation for Innovation (CFI) and Ontario Innovation Trust (OIT), and an NSERC student fellowship to the first author.

21 C293 Capturing and Holding Attention 21 Table 1 ANOVAs on Target Accuracy by Experiment and Task Experiment 1 Experiment 2 Word type Lag Wdtype x Lag Word type Lag Wdtype x Lag AB Task AB Task F 15.00*** 3.08* 2.78*** F 18.89*** 14.91*** 1.59 df 4, 84 5, , 420 df 4, 144 5, , 720 η η Capture Task Capture Task F 7.87*** F 8.24*** ** df 4, 84 5, , 420 df 4, 156 5, , 780 η η Note: * = p <.05; ** = p <.01, *** = p <.001

22 C293 Capturing and Holding Attention 22 Table 2 Correlations Among Ratings, First Target Recognition and Second Target Accuracy (Pearson r) Experiment 1 Experiment 2 AB Task Arousal Valence Extremity T2 Accuracy T1 Recogn AB Task Arousal Valence Extremity T2 Accuracy Corrected T2 Recogn Valence * -.16* -.04 Arousal.57*** -.42***.55***.34*** -.23*** Val Extr * T2 Acc -.57***.02.48*** *** Capture Task Arousal Valence Extremity Target Accuracy PT Recogn Capture Task Arousal Valence Extremity Valence * Arousal.57*** -.30**.52***.45*** Val Extr Targ Acc -.42*** Target Accuracy Corrected PT Recogn **.33***.15* -.17** -.27*** Note: * = p <.05; ** = p <.01, *** = p <.001

23 C293 Capturing and Holding Attention 23 Table 3 Simultaneous Regression Analyses of Valence Extremity and Arousal on Second Target Accuracy Predictors β R 2 F t-values Experiment 1 T2 Accuracy (AB) Complete Model Valence Extremity Arousal Target Accuracy (Capture) Complete Model Valence Extremity Arousal *** ** *** *** ** *** Experiment 2 T2 Accuracy (AB) Complete Model Valence Extremity Arousal *** *** *** Target Accuracy (Capture) Complete Model Valence Extremity Arousal *** *** *** Note: * = p <.05; ** = p <.01, *** = p <.001

24 C293 Capturing and Holding Attention 24 Table 4 Simultaneous Regression Analyses of Valence Extremity and Arousal on First Target Recognition Predictors β R 2 F t-values Experiment 1 T1 Recognition (AB) Complete Model Valence Extremity Arousal *** *** *** Pseudotarget Recognition (Capture) Complete Model Valence Extremity Arousal *** *** *** Experiment 2 Corrected T1 Recognition (AB) Complete Model Valence Extremity Arousal *** *** *** Corrected Pseudotarget Recognition (Capture) Complete Model Valence Extremity Arousal *** *** *** Note: * = p <.05; ** = p <.01, *** = p <.001

25 C293 Capturing and Holding Attention 25 Table 5 Simultaneous Regression Analyses of First Target Recognition and Arousal on Second Target Accuracy Predictors β R 2 F t-values Experiment 1 T2 Accuracy (AB) Complete Model T1 Memory Arousal *** *** Target Accuracy (Capture) Complete Model Pseudotarget Memory Arousal *** ** Experiment 2 T2 Accuracy (AB) Complete Model Corrected T1 Recognition Arousal *** *** Target Accuracy (Capture) Complete Model Corrected Pseudotarget Recognition Arousal *** ** * Note: * = p <.05; ** = p <.01, *** = p <.001

26 C293 Capturing and Holding Attention 26 Figure Captions Figure 1A. The sequence of stimulus events within a stream for the AB task, Experiments 1 and 2. Figure 1B. The sequence of stimulus events within a stream for the capture task, Experiments 1 and 2. Figure 2. Two-factor plot of the mean arousal and mean valence ratings for each of the 96 pseudo-target/t2 words (collapsed across participants) in Experiment 1. Figure 3A. The mean percentage of correct T2 responses as a function of T1 word-type and the lag between T1 and T2 in the AB task, Experiment 1. Error bars represent the standard error for each mean. Figure 3B. The mean percentage of correct target responses as a function of pseudo-target wordtype and the lag between pseudo-targets and targets in the capture task, Experiment 1. Error bars represent the standard error for each mean. Figure 4: Model showing the relationships among target accuracy, arousal ratings for the emotion word and recognition of the emotion word. The results suggest the relationship between arousal rating and target accuracy is mediated by encoding of the emotion words. Figure 5: Two-factor plot of the mean arousal and mean valence ratings for each of the 240 pseudo-target/t2 words (collapsed across the two participant groups), Experiment 2. Figure 6A: The mean percentage of correct T2 responses as a function of T1 word-type and the lag between T1 and T2 in the AB task, Experiment 2. Error bars represent the standard error for each mean. Figure 6B: The mean percentage of correct target responses as a function of pseudo-target wordtype and the lag between pseudo-targets and targets in the capture task, Experiment 2. Error bars represent the standard error for each mean.

27 C293 Capturing and Holding Attention 27 Fig. 1A Fig. 1B ms 500 ms 4 or 7 pre-target words 500 ms ms FLEW bitch T1 INPUT Arousing Negative Positive Neutral Distractor FLEW BITCH Pseudo-target INPUT Arousing Negative Positive Neutral Distractor Time MEAL 0-7 words between targets MEAL BLUE Neutral T2 (color word) GREEN Neutral target (color word) FAINT FAINT DISK 2-12 post-target words DISK 18 stimuli 18 stimuli AB Task Capture Task

28 C293 Capturing and Holding Attention 28 Fig. 2

29 C293 Capturing and Holding Attention 29 Fig. 3A % Correct T2 Report T1 - T2 Lag Negative Positive Taboo Neutral None

30 C293 Capturing and Holding Attention 30 Fig. 3B % Correct Target Report Negative Positive Taboo Neutral None Pseudotarget - Target Lag

31 C293 Capturing and Holding Attention 31 Fig. 4 Arousal Ratings Recognition Memory Target Accuracy

32 C293 Capturing and Holding Attention 32 Fig. 5